HAL Id: tel-03033790
https://tel.archives-ouvertes.fr/tel-03033790
Submitted on 1 Dec 2020
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Fei Gao
To cite this version:
Fei Gao. Three Essays on Marketing Interventions to Influence Consumer Judgments, Choices, and Behaviors. Business administration. HEC, 2020. English. �NNT : 2020EHEC0002�. �tel-03033790�
Institut Polytechnique de Paris 91120 Palaiseau, France cotutelle internationale de thèse
(Cadre à enlever)
Three Essays on Marketing Interventions to Influence Consumer Judgments, Choices, and Behaviors
Thèse de doctorat de l’Institut Polytechnique de Paris préparée à HEC Paris
École doctorale de l’Institut Polytechnique de Paris – (ED IP Paris) n°626
Spécialité de doctorat: Sciences de Gestion
Thèse présentée et soutenue à Jouy-en-Josas, le June 9th, 2020, par
GAO Fei
Composition du Jury :
Professor, University of Kansas (Psychology Department) Rapporteur Landau, Mark J.
Associate Professor, Universita' Bocconi (Marketing Department) Rapporteur Estes Zachary
Professor, ESSEC (Marketing Department) Président Laurent Gilles
Associate Professor, HEC Paris (Marketing Department) Examinateur Gu, Yangjie
Professor, HEC Paris (Marketing Department) Directeur de thèse Lowrey, Tina M.
Professor, HEC Paris (Marketing Department) Co-Directeur de thèse Shrum, L. J.
NNT : 2020EHEC0002
Institut Polytechnique de Paris 91120 Palaiseau, France
Titre : Trois essais sur les interventions marketing pour influencer les jugements, les choix et les comportements des consommateurs
Mots clés : Interventions marketing, comportements des consommateurs, psychologie du consommateur Résumé : Cette thèse se compose de trois essais qui
développent des interventions marketing pour influencer le jugement, le choix et les comportements des consommateurs. L'essai 1 étudie si, comment et quand les correspondances multimodales affectent les jugements en aval. Essay 2 développe une intervention comportementale visant à réduire le choix des consommateurs quant à la taille des portions de nourriture, qui peut être facilement utilisée dans les paramètres de commande de nourriture en ligne tels que les applications de livraison de nourriture. L'essai 3 étudie si le programme d'incitation prosocial peut motiver efficacement les consommateurs à participer à des programmes de référence en ligne.
Title : Three Essays on Marketing Interventions to Influence Consumer Judgments, Choices, and Behaviors Keywords : Marketing Interventions, Consumer Behaviors, Consumer Psychology
Abstract : This dissertation consists of three essays that develop marketing interventions to influence consumers judgment, choice, and behaviors. Essay 1 studies whether, how, and when crossmodal correspondences affect downstream judgments.
Essay 2 develops a behavioral intervention aimed at reducing consumers’ choices of food portion sizes, which can be easily used in the settings of online food ordering such as food delivery apps. Essay 3 studies whether prosocial incentive scheme can effectively motivate consumers to participate in online referral programs.
Three Essays on Marketing Interventions to Influence Consumer Judgments, Choices, and Behaviors
Fei Gao
A thesis submitted for the degree of
Doctor of Philosophy in Management with Specialization in Marketing at HEC Paris, France.
2020
Abstract
This dissertation consists of three essays that develop marketing interventions to
influence consumers judgment, choice, and behaviors. Essay 1 studies whether, how, and when crossmodal correspondences affect downstream judgments. Six experiments that investigated shape–taste correspondences show that whether the correspondences influence downstream taste judgments depends on the associative strength in memory between the shape and taste. The shape of a logo (angular vs. rounded) for a bottled water (sparkling, still) or the shape of the product itself (chocolate bar) influences judgments of the water’s carbonation and smoothness and the chocolate’s sweetness and bitterness. However, these effects occur only if the shape–
taste associative strengths reach a sufficient threshold (application threshold), even if the
correspondences have surpassed the activation threshold (statistical significance). The effects are not observable in very young children because they have not yet established shape–taste
correspondences, but the associative strengths increase with age, and eventually affect
downstream judgments. These processes are highly automatic, occurring even when visualization is impaired. Marketers can apply these findings to design packaging that will influence
consumers’ expectations and perceptions of product attributes.
Essay 2 develops a behavioral intervention aimed at reducing portion size choices based on conceptual metaphor theory. In six experiments, the authors show that the simple sequential presentation of two food images that move from partial to whole (e.g., a pizza with one vs. no missing pieces) reduces portion size choice compared to all other possible sequences. This effect occurs because the partial-to-whole sequence activates the metaphorical concept of fullness, which transfers to judgments of appetitive fullness (metaphor transfer effect). The effects occur
even when the sequential images are unrelated to food, is robust across languages, age groups, food type, and hypothetical and actual food choice contexts. The effect of image sequence on portion size choice is mediated by reduced perceptions of hunger and is attenuated when visualization abilities are inhibited. The intervention is conducive to implementation in online food-ordering contexts and applicable for both commercial and non-commercial situations (e.g., diet apps, school cafeteria apps).
Essay 3 studies whether prosocial incentive scheme can effectively motivate consumers to participate in online referral programs. Most online referral programs are based on promoting consumers’ self-interests, either by rewarding referrers only, or by rewarding both referrers and recipients. Very few of them employ a prosocial strategy which rewards recipients only (but not referrers). In this paper, we present five behavioral experiments (including a large-scale
randomized field experiment) in a variety of settings. We demonstrate the effectiveness of the prosocial strategy in motivating, not only referrers to make referrals, but also recipients to adopt the referred products. We show that the prosocial strategy can be as effective as the double-sided strategy, which rewards both referrers and recipients, albeit the prosocial strategy only requires half of the marketing expenditure for online retailers. We further find that referrers’ prosocial tendencies in the prosocial strategy are driven by altruism, while offering extrinsic rewards crowds out referrers’ altruistic motives. Interestingly, we also find that referrers’ altruistic tendencies in the prosocial strategy are inhibited with an increased action cost of referral behaviors, but not with an increased social cost of referral behaviors, thereby showing the asymmetrical moderating roles of action cost and social cost. We conclude the paper by offering managerial implications for designing effective prosocial reward strategies.
Abstrait
Cette thèse se compose de trois essais qui développent des interventions marketing pour influencer le jugement, le choix et les comportements des consommateurs. L'essai 1 étudie si, comment et quand les correspondances inter-modales affectent les jugements en aval. Six expériences qui ont étudié les correspondances forme-goût montrent que l'influence des correspondances sur les jugements de goût en aval dépend de la force associative dans la mémoire entre la forme et le goût. La forme d'un logo (angulaire ou arrondi) pour une eau en bouteille (pétillante, plate) ou la forme du produit lui-même (barre de chocolat) influence les jugements sur la carbonatation et la douceur de l'eau, ainsi que sur la douceur et l'amertume du chocolat. Cependant, ces effets ne se produisent que si les forces associatives forme-goût atteignent un seuil suffisant (seuil d'application), même si les correspondances ont dépassé le seuil d'activation (signification statistique). Les effets ne sont pas observables chez les très jeunes enfants car ils n'ont pas encore établi de correspondance forme-goût, mais les forces associatives augmentent avec l'âge, et finissent par affecter les jugements en aval. Ces processus sont hautement automatiques et se produisent même lorsque la visualisation est altérée. Les spécialistes du marketing peuvent appliquer ces résultats pour concevoir des emballages qui influenceront les attentes des consommateurs et leur perception des attributs du produit.
L'essai 2 développe une intervention comportementale visant à réduire les choix de taille de portion basée sur la théorie de la métaphore conceptuelle. Dans six expériences, les auteurs montrent que la simple présentation séquentielle de deux images d'aliments qui passent de partiel à entier (par exemple, une pizza avec une ou aucune pièce manquante) réduit le choix de la taille des portions par rapport à toutes les autres séquences possibles. Cet effet se produit parce que la
séquence partielle à entière active le concept métaphorique de plénitude, qui se transforme en jugements de plénitude appétitive (effet de transfert de métaphore). Les effets se produisent même lorsque les images séquentielles ne sont pas liées à la nourriture, sont robustes à travers les langues, les groupes d'âge, les types d'aliments et les contextes de choix alimentaires
hypothétiques et réels. L'effet de la séquence d'images sur le choix de la taille des portions est médié par des perceptions réduites de la faim et est atténué lorsque les capacités de visualisation sont inhibées. L'intervention est propice à la mise en œuvre dans des contextes de commande de nourriture en ligne et s'applique à la fois aux situations commerciales et non commerciales (par exemple, applications de régime, applications de cafétéria scolaire).
L'essai 3 étudie si un programme d'incitation prosocial peut motiver efficacement les consommateurs à participer à des programmes de référence en ligne. La plupart des programmes de parrainage en ligne sont basés sur la promotion des intérêts personnels des consommateurs, soit en récompensant uniquement les référents, soit en récompensant à la fois les référents et les destinataires. Très peu d'entre eux utilisent une stratégie prosociale qui récompense uniquement les destinataires (mais pas les référents). Dans cet article, nous présentons cinq expériences comportementales (y compris une expérience de terrain randomisée à grande échelle) dans divers contextes. Nous démontrons l'efficacité de la stratégie prosociale pour motiver non seulement les référents à faire des références, mais également les destinataires à adopter les produits référés.
Nous montrons que la stratégie prosociale peut être aussi efficace que la stratégie double face, qui récompense à la fois les référents et les destinataires, même si la stratégie prosociale ne nécessite que la moitié des dépenses marketing des e-commerçants. Nous constatons en outre que les tendances prosociales des référents dans la stratégie prosociale sont motivées par l'altruisme, tout en offrant des récompenses extrinsèques évince les motivations altruistes des
référents. Fait intéressant, nous constatons également que les tendances altruistes des référents dans la stratégie prosociale sont inhibées avec un coût d'action accru des comportements
d'orientation, mais pas avec un coût social accru des comportements d'orientation, montrant ainsi les rôles modérateurs asymétriques du coût d'action et du coût social. Nous concluons l'article en proposant des implications managériales pour la conception de stratégies de récompense
prosociales efficaces.
Acknowledgements
My deep gratitude goes first to my supervisors, Professor Tina M. Lowrey and Professor L. J. Shrum. I appreciate you both for giving me an opportunity to be your student. I am
sincerely grateful for your expertly guidance. Particularly your unwavering enthusiasm,
optimism, and encouragements always make me feel reassured and confident about my research.
I completely enjoy the countless intellectual interactions with you both. It was truly my honor and my fortune to work with you during these unforgettable PhD years. Professor Xitong Li, my co-author and my friend, I want to send my huge appreciation to you for everything you have done for me. You were always beside me during the happy and hard moments to motivate me, for which I am very grateful. You provided me with numerous great opportunities and resources that are rare for PhD students to get to explore my research ideas. My huge thanks also to Professor Kristine De Valck. It is really my luck that I have you during my PhD years. Your extreme warmth and empathy help me get through countless difficult moments in the past six years. You are my mentor and my friend, but I prefer to treat you as my family member. My big thanks next extend to Melanie, Françoise, Britta, Julie, Christelle, and Francine. Without you all, I cannot imagine what my PhD life would be like. I really appreciate your friendliness and patience when I need your help. I also really appreciate all the help from Peter, Cathy, Anne- Laure, Daniel, Marc, Anne-Sophie, Yangjie, Andreas, Valeria, Ludovic, and Dominique during the past six years. Thank you so much to all my good friends accompanying me during my PhD years: Fan, Sukhyun (my hyung), Zilu, Yapei, Jing, Yufei, and Ana. Finally, I acknowledge the people who mean a lot to me: Mom and Dad, thank you for your unconditional love and support in my life, and giving me liberty to choose what I desired; my parents in law, thank you for your moral support and love; my wife, Weiwei, words would never say how grateful I am to you for your continued and unfailing love, support, and understanding during my pursuit of PhD. I consider myself the luckiest in the world to have such a lovely family.
Essay 1: Altering Taste Judgments with Shapes: How and When Shape-Taste Crossmodal Correspondences Can Be Applied in Marketing Designs
ABSTRACT
Crossmodal sensory correspondences (associations between two different sensory modalities) are well-established, but whether, how, and when these correspondences affect downstream
judgments are not. Six experiments that investigated shape–taste correspondences show that whether the correspondences influence downstream taste judgments depends on the associative strength in memory between the shape and taste. The shape of a logo (angular vs. rounded) for a bottled water (sparkling, still) or the shape of the product itself (chocolate bar) influences
judgments of the water’s carbonation and smoothness and the chocolate’s sweetness and bitterness. However, these effects occur only if the shape–taste associative strengths reach a sufficient threshold (application threshold), even if the correspondences have surpassed the activation threshold (statistical significance). The effects are not observable in very young children because they have not yet established shape–taste correspondences, but the associative strengths increase with age, and eventually affect downstream judgments. These processes are highly automatic, occurring even when visualization is impaired. Marketers can apply these findings to design packaging that will influence consumers’ expectations and perceptions of product attributes.
Keywords: shape, taste, crossmodal correspondence, crossmodal downstream effect, sensory marketing, memory
When consumers make choices about which foods to buy, they rely on their existing taste preferences, and these preferences can vary across consumers. For example, some consumers prefer sweeter chocolates, some prefer more bitter chocolates. Some consumers like still waters, some like sparkling waters. Even within the latter category, tastes can vary. Some consumers prefer highly carbonated waters, others less carbonated.
However, when choosing a food for the first time, it is difficult if not impossible for consumers to know what the food tastes like before purchase. Particularly for in-store purchases, the packaging may provide little or no explicit information. Moreover, using the examples of sparkling water and chocolates, even when explicit information is available, it may provide little help to non-connoisseurs (i.e., interpreting carbonation in terms of mg/l or pH, or chocolates in terms of percent cocoa). In such situations, how do consumers decide which product is most likely to meet their taste needs?
Building on research on crossmodal sensory correspondences (matches between features in one sensory modality and features in another sensory modality; Spence 2012), we propose that consumers may rely on subtle shape cues that signal to them the likelihood of a specific taste, such as the shape of a product’s logo or the shape of the product itself. In fact, crossmodal correspondences among all five senses are well-established (Spence 2012). Recent research, moreover, suggests that these associations can affect downstream product judgments. For
example, the shape of a logo influences haptic expectations for a product (Jiang et al. 2016), and the color saturation of a product (Hagtvedt and Brasel 2017) and the sounds of brand names (Lowrey and Shrum 2007) influence size expectations. These expectations, in turn, influence attitudes and behaviors.
This general set of findings has obvious appeal for marketers. Sensory-related attributes of a product—the shape of a logo, the color of a product, the sound of a brand name—are easy to manipulate, particularly with new product introductions. Consequently, a) given the plethora of research documenting crossmodal correspondences, and b) the growing body of research showing downstream crossmodal effects on product judgments, marketers might reasonably assume that all they need to leverage these effects is knowledge of the established crossmodal correspondences.
In the present research, however, we show that this assumption is wrong. In six experiments, we show that whether crossmodal correspondences produce the expected downstream effects depends on qualities of the crossmodal correspondences. In particular, we demonstrate that downstream effects are a function of the strength of association in memory between the two sensory features, and that merely demonstrating significant crossmodal correspondences (those occurring at a rate greater than chance; reaching activation threshold) is a necessary but not sufficient condition for realizing downstream effects. Rather, the associative strength in memory must reach a certain threshold (application threshold) to influence downstream judgments.
Indeed, as we detail in the following section, although the existence of shape–taste correspondences has been consistently documented, research on the effects of these
correspondences on downstream judgments has been inconsistent. Our research provides a potential explanation for these inconsistencies.
We also investigate the automaticity of the downstream effects, and in particular whether they require a visualization process or are mental-load-independent. Whether the process is dependent on visualization is not merely a theoretical issue, it also has important implications for the application of shape–taste correspondence downstream effects in marketing
settings. If the generation of the downstream effects requires a visualization process that
consumes substantial mental resources, the crossmodal effects will likely be severely diminished in both online and offline shopping environments that distract visual attention and impede visualization, such as simultaneous visual displays of various products in a limited space.
Our research makes several important contributions. First, we contribute to the growing literature on the effects of crossmodal correspondences on consumer judgments by
demonstrating the conditions under which these downstream effects occur. Second, we contribute to the literature on the processes underlying crossmodal correspondence effects by addressing the extent to which the process is automatic, and particularly whether visualization is required for downstream effects. Research has been inconsistent on the need for visualization, and our research provides a potential explanation for the conflicting findings. Third, we
contribute to the developmental literature on crossmodal correspondence effects, and show that, at least for the shape–taste correspondences we investigate, the strength of the crossmodal associations are not detectable in very young children, but develop with age, eventually reaching a sufficient strength to produce the same downstream effects noted for adults. Finally, each of these theoretical contributions has direct and important managerial implications that can easily be implemented.
THEORETICAL DEVELOPMENT
Shape is one of the most important properties of objects in nature. In marketing practice, shape is an important visual element of art used in advertising, branding, and packaging design (Krishna, Cian, and Aydınoğlu 2017). Humans not only perceive shapes, but also develop associations between shapes and other sensory features (crossmodal associations). For example,
people reliably match shapes with features of sensory modalities such as color (Chen, Tanaka, and Watanabe 2015), pitch (Walker 2012), and tactile hardness (Lundholm 1921).
Shape and Taste Correspondences and Their Downstream Effects
Consumers can reliably match shapes and tastes. For example, angular shapes are associated with bitter, sharp, and carbonated tastes, and rounded shapes are associated with sweet, mild, and smooth tastes (Ngo, Misra, and Spence 2011; Spence et al. 2013; Spence and Gallace 2011).
However, whether the particular shapes influence corresponding downstream taste judgments is less clear. For example, in one study with British participants, beverages in bottles with an angular shape were expected to taste more sour and less sweet than beverages in bottles with a rounded shape (Velasco et al. 2014). In contrast, in another study, British participants thought foods served on round plates tasted sweeter than those same foods served on angular (square) plates, but plate shape did not affect perceptions of sourness (Fairhurst et al. 2015). Still other studies found no effects of plate shape (angular vs. rounded) on Spanish participants’
(Piqueras-Fiszma et al. 2012) and Canadian participants’ (Stewart and Goss 2013) perceptions of sweetness. Similar inconsistent findings on shape–taste crossmodal effects have been reported for other tastes, such as sharpness (Harrar and Spence 2013). Thus, although consumers can reliably match particular shapes to tastes, downstream crossmodal effects are not consistently observed.
Although it is difficult to generalize across a small number of studies that differ on many characteristics, as we discuss in the following sections, shape–taste correspondences are likely learned associations. If so, the strengths of these associations may vary across learning
environments and even vary in the course of individual development. Consequently, we propose that even though shape–taste associations may exist for particular people, the associations may not be sufficiently strong to produce downstream effects on taste judgments. Thus, the existence of shape–taste associations may be a necessary but not sufficient condition for producing
downstream effects.
Underlying Processes
Semantic associative strength. We conceptualize crossmodal sensory correspondences in terms of a spreading-activation theory of semantic processing (Collins and Loftus 1975).
Spreading-activation theory regards human memory as a semantic associative network in which concepts and knowledge are stored in the form of nodes. Associations are represented by
connections between the nodes (e.g., the nodes of shapes and tastes), and the length of the connections between nodes (concepts) reflects the strength of their association: the shorter the linkage between nodes, the stronger the association between them. Thus, in terms of shape–taste correspondences, when a shape node is activated (e.g., angular shape), it will automatically spread out to connected nodes (e.g., bitterness, sourness, carbonation), which in turn may be activated. However, a simple linkage between nodes does not guarantee that activation of one node will activate another linked node. Rather, for the association (link) to be activated, the associative strength must reach a sufficient threshold (Collins and Loftus 1975). We refer to this threshold as the activation threshold.
Further, the issue of associative strength thresholds also has important implications for whether the crossmodal association has downstream effects on taste judgements. For these associations to impact downstream judgments, we argue that the crossmodal association must be
sufficiently strong that it is not only activated, but also spontaneously applied to the downstream judgment. In other words, we argue that the associative strength must reach a sufficient threshold for automatic application to downstream judgments. We refer to this threshold as the application threshold. Moreover, the application threshold must necessarily be greater than the minimum activation threshold. We test this proposition in our studies. If our reasoning is correct, it may potentially explain why research on known shape–taste correspondences does not always produce downstream effects on taste judgments.
Symmetry of shape–taste associations. Shape–taste correspondences (and crossmodal correspondences in general) are typically conceptualized and assessed by comparing polar opposites on the two dimensions (shape and taste). For example, angular shapes are more
associated with bitter than with sweet tastes, and rounded shapes are more associated with sweet than with bitter tastes (Ngo et al. 2011). Thus, shape–taste correspondences may be assessed by first asking participants to either imagine or taste a food or drink, and then mark a point along a visual analog scale anchored at one end by an angular shape and at the other end by a rounded shape. Support for the crossmodal correspondence is shown when the expected congruent ratings (e.g., angular with bitter, rounded with sweet) differ from the midpoint of the scale, and from each other (cf. Ngo et al. 2011; Spence and Gallace 2011).
Although these tests are appropriate for assessing the existence of a crossmodal
correspondence, it is often difficult to discern whether the particular correspondences between shape and taste have similar associative strengths. For example, it may be that the association between angular shapes and bitter tastes is stronger than the corresponding associations of the opposing shapes and tastes (e.g., rounded shapes with sweet tastes). If so, then the relative associative strength may influence the extent to which these associations influence downstream
taste judgments. Consequently, it is possible that, using the examples just described, angular shapes may influence taste judgments of bitterness, but round shapes may not influence taste judgments of sweetness. Our research also explores this possibility.
Shape–taste associations develop over time. Given research showing cross-cultural differences in shape–taste correspondences (Bremner et al. 2013), we assume the
correspondences are likely learned over time from environmental cues. If so, they should be a function of human developmental processes, which suggests that the effects may not be observed at early stages of child development. In fact, there is a close relationship between the
development of semantic associations and abstract reasoning ability (Hutchison 2003). Building on Piaget’s (1964) theory of cognitive development, John (1999) suggests that children in the perceptual stage (ages 3-7) focus predominantly on perceptual features of stimuli, typically a single dimension, but lack abstract reasoning that connects multiple features. However, as children move into the analytical stage (ages 7-11), their thinking shifts from a perceptual orientation to a more abstract orientation, which enables them to consider abstract, common features of objects and events, and thus develop corresponding semantic associations. Thus, we expect that children are more likely to develop shape–taste correspondences that reach the activation threshold at the analytical stage. Further, these associative strengths may further increase with age, to the point that they reach the application threshold, and thus the effects of shape–taste correspondences on downstream judgments are also likely to start occurring at this stage. If so, crossmodal effects on downstream judgments are more likely to occur among adults and older children than among younger children.
Automaticity of shape–taste correspondence downstream effects. Whether the processes underlying crossmodal correspondence downstream effects are automatic is not only an
important theoretical question, it also has important implications for the efficacy of marketing applications. Addressing this question requires two considerations. The first is defining automaticity. Although there is not full consensus on the definition, recent conceptualizations consider automaticity as a function of at least four features: they are fast (i.e., occur very early in information processing), occur outside of awareness, are uncontrollable, and are mental load- insensitive (Moors and De Houwer 2006). Thus, rather than viewing automaticity as a unitary or all-or-none construct, automaticity may be diagnosed by the extent to which a process exhibits these features.
The second consideration is the delineation of two different types of crossmodal
correspondence effects: the automaticity of the associations themselves, and the automaticity of the process of applying the correspondences to downstream judgments. For example,
automaticity of a crossmodal association between shape and taste pertains to the extent to which presentation of one sensory dismission (e.g., an angular shape) automatically activates the presumed associated taste dimension (bitter). The general consensus is that most crossmodal associations are quite automatic (Spence and Deroy 2013a).
With respect to the automaticity of the application of associations to downstream judgments, the findings are less clear. Recent research in marketing provides some clear examples. In the context of the effects of sound symbolism of brand names on taste attributes, Yorkston and Menon (2004) found that the crossmodal effects of sounds on taste judgments occurred outside of awareness and were mental-load-independent. In contrast, the crossmodal effects of shape on haptic judgments does not appear to be mental-load-independent, but instead requires a visualization process that consumes substantial mental resources. For example, in one study, the shape of a logo (angular, rounded) influenced perceptions of comfort and durability of
a sofa (Jiang et al. 2016). However, these effects were eliminated under visual load conditions that inhibit the visuospatial working memory, and the effects were observed only for those scoring higher on an imagery disposition measure. Similar findings were reported by Lowe and Haws (2017), who found that the crossmodal downstream effects of sound pitch on perceived product size were driven by a visualization process, and thus were eliminated under visual load conditions.
One potential explanation for these seemingly conflicting findings is that automaticity may be a function of how individuals construct sensory judgments of the target product’s
attributes. More specifically, whether crossmodal effects require a visualization process is likely to be contingent on whether individuals spontaneously generate visual images of the target when they make judgments about its sensory qualities, such as whether a sofa will be comfortable or durable (Jiang et al. 2016) or the size of a hamburger (Lowe and Haws 2017). In fact, mental imagery is usually involved when making perceptual judgments of haptic features (James et al.
2002), consistent with the findings of Jiang et al. (2016), and individuals typically spontaneously generate mental imagery when they judge visual features of objects such as size (Spence and Deroy 2013b), consistent with the findings of Lowe and Haws (2017). In such cases, if the visualization process is inhibited, then the application of the accessible crossmodal associations may also be inhibited.
However, not all sensory judgments require visualization. One example is taste
judgments. Mental imagery is not a necessary component for taste judgments (Kobayashi et al.
2004), and thus it is unlikely that individuals would spontaneously generate images to judge how a product should taste. In such cases, the accessible crossmodal association (e.g., shape–taste)
should be successfully applied to taste judgments even if visual processing abilities are constrained. Our research also addresses these possibilities.
HYPOTHESES AND EMPIRICAL OVERVIEW
The primary hypothesis we test is whether shape–taste correspondences influence downstream judgments depends on the associative strength in memory between the two sensory features. We expect that the effect of shape on taste judgments will be moderated by associative strength, such that stronger associations will produce stronger downstream effects. Moreover, we expect that even when shape–taste associations are statistically significant (reach activation threshold), they will not affect downstream judgments unless the associative strength reaches a sufficient threshold (application threshold).
Studies 1A-1C tested the primary hypothesis by focusing on the downstream effects of two pairs of primary shape–taste correspondences: the relations between angular and round shapes on judgments of carbonation and smoothness in bottled waters, and the relations between angular and round shapes on judgments of bitterness and sweetness of chocolate bars. For the bottled water studies in particular, we conducted the studies with participants from a country (France) in which bottled water is a highly differentiated product category, and personal tastes exhibit clear variation in terms of preferences for sparkling and still waters and degrees of carbonation and smoothness. For the bottled water studies, we manipulated shape via the bottle logo, and for the chocolate studies, we manipulated the product shape.
In Study 2, we tested a naturally occurring boundary condition regarding associative strength. We expect that if associations between shape and taste are developed over time, then the associations should occur after children enter the analytical stage of cognitive development in
which their abstract ability is developed, and should be stronger for older children than for younger children, which should in turn have differential effects on taste judgments as a function of age.
Finally, we tested whether downstream effects of shape–taste correspondences require a visualization process (Studies 3A and 3B). Because judgments of taste typically do not require a visualization process, we expect that crossmodal shape–taste effects should be observed even when a visualization process is impaired (Study 3A). However, if consumers are conditioned to visualize tastes, then impairing visualization during the taste judgment process should reduce or eliminate the crossmodal effects (Study 3B).
Across all studies, we analyzed the data only after all responses had been collected, no participants’ data were excluded from analyses, and details of all measures and manipulations are provided in the Web Appendix.
STUDY 1A: LOGO SHAPE AND TASTE EXPECTATIONS FOR BOTTLED WATER Study 1A tested whether shape–taste correspondences influence downstream judgments on product taste attributes and whether the generation of the downstream effects depends on the associative strength of the shape–taste correspondences. We manipulated the logo shape of a bottled drink (sparkling or still water) and compared participants’ judgments of how carbonated and smooth the drinks would taste between the logo conditions. We also measured participants’
associative strengths of angular–carbonated and rounded–smooth correspondences and tested whether the downstream effects occurred only if the associative strengths reached a sufficient threshold (application threshold).
Method
Participants and design. Two hundred forty-three French college students (144 women;
Mage = 22.03 years) participated in the study in exchange for an opportunity to participate in a lottery to win 30€. Participants were randomly assigned to conditions in a 3 (logo shape: angular, rounded, no logo) × 2 (drink type: sparkling, still water) × 2 (taste judgment: carbonated,
smooth) mixed design, with logo shape and drink type as between-subjects factors and taste judgment as a within-subjects factor.
Procedure. Participants were informed that a beverage company was launching a new bottled drink and wanted consumers’ feedback on the packaging. Next, we showed participants a bottle that had either an angular or rounded logo, or no logo, with all other package elements held constant (see Web Appendix A). Participants then rated how carbonated and smooth they thought the drink would taste (1 = not at all, 9 = very much), and then (except for those in the control condition) indicated their attitudes toward the bottle logo along a 3-item, 9-point scale (1
= not at all, 9 = very; α = .94). Finally, to measure associative strength, we showed participants two slider line scales, anchored by an angular shape and a rounded shape at the left (-4) and right end (4). The shapes were the ones used in the classic “Kiki” and Bouba” studies (Ramachandran and Hubbard 2001). Participants dragged the slider to a point that best matched their taste experience when drinking sparkling or still water.
Results and Discussion
A 2 (logo shape) × 2 (drink type) ANOVA confirmed that there was no difference on participants’ liking of the logos across conditions (ps > .79).
Shape–taste crossmodal correspondences. We first assessed the presence of crossmodal correspondences of angular shape–carbonation and rounded shape–smooth, and the relative strengths of these associations. To do so, we used participants’ slider scale ratings of the shape–
taste correspondences, and conducted separate one-sample t-tests with 0 (i.e., the mid-point of the scale) as the test value to examine the associations between angular and rounded shapes and tastes sparkling and still water. As expected, the associations between angular shapes and the carbonated taste of sparkling water (M = -2.98, SD = 1.31, t(242) = -35.39, p < .001), and
between rounded shapes and the smooth taste of still water (M = 2.07, SD = 1.92, t(242) = 16.75, p < .001), were both significant. To determine the relative associative strength between angular shape–carbonation and rounded shape–smoothness, we compared the distances from
participants’ taste ratings to the mid-point between the two congruent conditions (the longer the distance, the stronger the associative strength). Paired-samples t-test indicate that participants’
angular–carbonation associative strength (M = 2.98, SD = 1.31) was significantly stronger than the rounded–smooth associate strength (M = 2.07, SD = 1.92, t(242) = 7.11, p < .001).
Shape–taste crossmodal effects. We next examined whether the shape–taste
correspondences influenced participants’ judgments of the tastes of sparkling and still bottled waters. To do so, we conducted separate 3 (logo shape: angular, rounded, no logo) × 2 (taste judgment: carbonation, smoothness) mixed-model ANOVAs, with logo shape as a between- subjects factor and taste judgment as a within-subjects factor, for sparkling and for still water.
The results of these analyses can be seen in Figure 1. (Note that although our primary interest is the effect of logo shape on judgments of carbonation for sparkling water (top left panel of Figure 1), and the effect of logo shape on judgments of smoothness for still water (bottom right panel of Figure 1), given the fully crossed design, participants also rated the smoothness of the sparkling
water and the carbonation of the still water. Although these latter judgments are not central to our main concerns, they are valuable for checking the discriminant validity of our
measurements.)
For sparkling water, only the interaction of logo shape and taste judgment was significant (F(2, 118) = 9.40, p < .001). As the left portion of Figure 1 (top panel) shows, consistent with our expectations, participants rated the sparkling water as more carbonated in the angular logo condition (Mangular = 5.40, SD = 2.18) than in the no logo condition (Mno logo = 4.07, SD = 1.74, t(118) = 3.02, p = .003) and in the rounded condition (Mrounded = 4.28, SD = 2.00, t(118) = 2.54, p
= .01), with no difference between the latter two conditions (t(118) = .46, p = .65). These results suggest that the angular logo increased expectations of the sparkling water’s carbonation, but the rounded logo had no effect.
For still water, the main effect of taste judgment (F(1, 119) = 337.93, p < .001) was significant, as was the interaction of logo shape and taste judgment (F(2, 119) = 4.75, p = .01).
As the bottom right portion of Figure 1 indicates, although participants rated the still water as more smooth in the rounded logo condition (M = 6.38, SD = 2.08) than in the angular logo condition (M = 5.05, SD = 2.09; t(119) = 2.86, p = .005), their ratings of smoothness for still water did not differ between the rounded (M = 6.38, SD = 2.08) and no logo conditions (M = 6.41, SD = 2.10; t(119) = .09, p = .93), but ratings of smoothness decreased in the angular logo condition compared to the no logo condition (t(119) = 2.96, p = .004). These findings indicate that the rounded logo had no effect on judgments of smoothness in the still water, but that the angular logo decreased expectations of a smooth taste.
The results of the other conditions are also instructive. Although smoothness is not an obvious dimension on which to rate sparkling water, participants nevertheless logically rated the
sparkling water as less smooth in the angular logo condition (Mangular = 4.15, SD = 1.87) than in the rounded (Mrounded = 5.68, SD = 2.12, t(118) = 3.25, p = .002) and no logo conditions (Mno-logo
= 5.38, SD = 2.27, t(118) = 2.63, p = .01), with no difference between the latter two conditions (t(118) = .65, p = .52; Figure 1, top right panel). These results again suggest that the effect of logo shapes on judgments of smoothness are driven by the angular logo, with no effect of the rounded logo, which does not differ from the control group. For judgments of the expected level of carbonation for a still water, participants’ ratings of carbonation for the still water were at a very low level and did not differ between the logo conditions (Mangular = 1.85, SD = .88, Mrounded
= 1.73, SD = .99, Mno-logo = 1.71, SD = 1.03, F(2, 119) = .28, p = .76; Figure 1, bottom left
panel), which is consistent with the features of still water and suggests good discriminant validity of our measures.
[INSERT FIGURE 1 HERE]
Moderating effects of associative strength. The results just discussed indicate that the angular logo affects perceptions of carbonation, but the rounded logo has no effect on
perceptions of smoothness, even though the associative strengths for both were significant. These results suggest that the associative strength of the angular–carbonated correspondence reached the application threshold, but that the rounded–smoothness association did not. To further explore these possibilities, and to specifically test the hypothesis that associative strength moderates the effect of shape–taste correspondences on downstream taste judgments, we conducted separate moderation analyses for sparkling and still water using Hayes’ (2013) PROCESS Model 1, with logo shape (angular or rounded vs. no logo) as independent variables, taste judgment (carbonation, smoothness) as dependent variables, and associative strength (angular/carbonated or rounded/smooth) as moderators. As expected, the interactions between
shape and associative strength for both sparkling (B = -.74, SE = .33, t = -2.26, p = .026) and still water (B = .82, SE = .23, t = 3.61, p < .001) were significant, indicating that the associative strength moderates the crossmodal effect of shapes on downstream taste judgments.
To examine the application threshold, we conducted a floodlight analysis (Spiller et al.
2013). The Johnson-Neyman technique (Johnson and Neyman 1936) indicated that for sparkling water, the point at which the effects become significant is an associative strength of -2.56. For still water, the associative strength threshold was 3.88. Thus, at an aggregate level, given that participants’ associative strength of angular shape and carbonation (M = -2.98) was beyond the threshold (- 2.56), angular logos significantly enhanced carbonation judgements of sparkling water. However, given that participants’ associative strength between rounded shape and smoothness (M = 2.07) was far below the threshold (3.88), rounded logos did not enhance smoothness judgments of still water.
Supplementary studies: Replications. In the current study, we used an explicit method (visual analog scale) for measuring associative strength. Although the results were consistent with our hypotheses, other methods (e.g., implicit methods) are likely more precise and thus better suited for measuring associative strength. Given the importance of associative strength in our theorizing (and particularly the differences in associative strengths for conceptually similar shape–taste matching relationships), we conducted two additional studies to demonstrate the generalizability of our findings. (We provide full details of methods and results in Web Appendix A).
In the first study, we measured associative strength with the Implicit Association Test (IAT). The experiment used a 2 (shape: angular, rounded) × 2 (shape/taste congruity: congruent, incongruent) within-subjects design, and the results replicated the findings of Study 1A.
Significant associations were observed between both angular shapes and carbonation, and between rounded shapes and smoothness (response latencies were shorter in the congruent groups than in the incongruent groups for both angular and rounded shapes). However, the strength of association was again significantly stronger for the angular–carbonated association than for the rounded–smooth association (response latencies in the congruent group of angular shape were shorter than that of rounded shape).
The second study measured associative strength using a lexical decision task in which participants were first exposed to either angular or rounded shapes, and we then measured how quickly participants recognized either a sparkling or still water. Again, our findings were confirmed. Angular shapes facilitated participants’ responses to sparkling water (shorter response latencies than for rounded shapes), but rounded shapes did not facilitate participants’
responses to still water (response latencies did not differ from exposure to angular shapes). These results are consistent with the findings of shape–taste correspondence effects at an explicit
judgmental level, namely, that the priming effects of angular shape on carbonation and rounded shape on smoothness are asymmetrical.
Taken together, the current study provides evidence that conceptually symmetrical shape- –taste correspondences (angular–carbonated vs. rounded–smooth) may have different
(asymmetrical) associative strengths. More importantly, we show that the associative strengths determine whether the crossmodal associations will influence downstream taste judgments. We provide direct evidence (via moderation analysis) that shape–taste correspondences produce downstream effects only when the associative strengths reach a sufficient (application) threshold, and this threshold is greater than the activation threshold. In the next two studies, we test the generalizability of these effects.
STUDY 1B: ASSYMETRICAL CROSSMODAL EFFECTS ON TASTE PERCEPTIONS The objective of Study 1B was to demonstrate the generalizability of the findings of Study 1A by replicating the effects with a different downstream judgment (actual taste
perceptions following a taste test) and using a different manipulation of logo shape. Based on the results of Study 1A, we expected that the angular logo would influence perceptions of
carbonation, but that the rounded logo would not influence perceptions of smoothness.
Method
Participants and design. Two hundred ninety-three French college students (176 women;
Mage = 22.72 years) who participated in return for 2€ were randomly assigned to conditions in a 3 (logo shape: angular, rounded, no logo) × 2 (drink type: sparkling, still water) × 2 (taste
judgment: carbonated, smooth) mixed design, with logo shape and drink type as between- subjects factors and taste judgment as a within-subjects factor.
Procedure. In a study ostensibly to test a new bottled drink, participants were told that the product was in an early stage of testing, and it did not yet have a brand name, just a rough logo (in the logo conditions only), and that we were only interested in their taste ratings. Participants were individually tested in a room that had been set up in advance. (See Web Appendix B for precise details of procedures, manipulations, and measures). Each participant was given a bottle of either sparkling or still water, with either an angular logo, rounded logo, or no logo. Just prior to their entry into the room, the bottle had been filled with a popular brand of either still or sparkling water. Participants were provided written instructions indicating that we wanted them to taste the product and then answer some questions about it. Next, participants tasted the water
and rated how carbonated and smooth they thought the drink tasted, each along a 16 cm line scale (anchors: not at all, very). (Numbers in results section indicate length in cm, range 0-16).
Finally, participants indicated their attitudes toward the logo (except in the no-logo condition) with the same 3-item scale used in the previous experiment (α = .93).
Results and Discussion
A 2 (logo shape) × 2 (drink type) ANOVA confirmed that there were no differences in participants’ liking of the logos across conditions (ps > .42).
Shape–taste crossmodal effects. To test whether logo shape influenced taste perceptions, we conducted separate 3 (logo shape: angular, rounded, no logo) × 2 (taste judgment:
carbonation, smoothness) mixed-model ANOVAs, with logo shape as a between-subjects factor and taste judgment as a within-subjects factor, for sparkling and for still water.
For sparkling water, the main effect of taste judgment was significant (F(1, 143) = 85.42, p < .001), as was the interaction of logo shape and taste judgment (F(2, 143) = 5.13, p = .007).
As expected, participants thought the sparkling water tasted more carbonated when the bottle had an angular logo (Mangular = 13.48, SD = 2.44) than when the bottle (with the exact same sparkling water) had no logo (Mno logo = 12.33, SD = 2.64, t(143) = 2.14, p = .034) or had a rounded logo (Mrounded = 12.17, SD = 2.89, t(143) = 2.42, p = .017). The latter two conditions did not differ (t(143) = -.30, p = .77). These results suggest that the angular logo increased perceptions of how carbonated the water tasted, but that the rounded logo had no effect, replicating Study 1A.
For still water, the main effect of taste judgment (F(1, 144) = 1462.23, p < .001) and the interaction (F(2, 144) = 3.43, p = .035) were significant. As with Study 1A, participants thought the still water tasted more smooth when the bottle had a rounded logo (M = 12.90, SD = 2.88)
than when it had an angular logo (M = 11.34, SD = 3.99; t(144) = 2.27, p = .025). However, participants’ perceptions of smooth taste did not differ between the rounded logo (M = 12.90, SD
= 2.88) and no logo conditions (M = 12.81, SD = 3.24; t(144) = .13, p = .89), but they rated smoothness as less in the angular than in the no logo conditions (t(144) = 2.13, p = .035). These results suggest that the rounded logo had no effect on perceptions of smoothness in the still water, but the angular logo reduced perceptions of smoothness, and also replicate the findings of Study 1A.
Also consistent with Study 1A, participants thought the sparkling water tasted less smooth in the angular logo condition (Mangular = 7.04, SD = 3.82) than in the rounded (Mrounded = 8.71, SD = 3.75, t(143) = -2.04, p = .04) and no logo conditions (Mno-logo = 9.24, SD = 4.47, t(143) = -2.70, p = .008), with no difference between the latter two conditions (t(143) = -.65, p
= .52). For the ratings of carbonation of the still water, they were again very low, and did not differ between logo conditions (Mangular = .67, SD = 1.88, Mrounded = .52, SD = 1.26, Mno-logo = .41, SD = 1.07, F(2, 144) = .40, p = .67).
The results of Study 1B again confirm that logo shape influences downstream taste judgments, but that the downstream effects as a function of angular versus rounded logos are asymmetrical. As in Study 1A, the effects of the angular logo on carbonated taste were larger than the effects of the rounded logo on smooth taste, and also like Study 1A, the rounded logo did not have a significant effect on taste judgments. The results also show that the effects of logo shape not only asymmetrically affect product expectations, but also actual perceptions of the product’s taste, even though the particular sparkling or still water tasted was the same across logo conditions.
STUDY 1C: ASSYMETRICAL CROSSMODAL EFFECTS: CHOCOLATE BARS The objective of Study 1C was to test whether the observed effects generalize to a different product category and corresponding taste judgments. We used chocolate bars as the product category, and taste judgments pertained to expectations of sweetness or bitterness. Shape was manipulated via the shape of the product itself (angular vs. rounded). We expect that angular shapes will be associated with more bitter tastes, and rounded shapes will be associated with sweeter tastes (Spence and Gallace 2011). Further, as with Study 1A, we again directly measured associative strength of the shape–taste correspondences, which we expect will moderate the effects of these correspondences on downstream taste judgments.
Method
Participants and design. One hundred UK-based participants (50 women; Mage = 39.50 years) who were recruited from the Prolific online panel in return for a nominal fee were randomly assigned to conditions in a 2 (product shape: angular, rounded) × 2 (taste judgment:
bitter, sweet) mixed design, with product shape as a between-subjects factor and taste judgment as a within-subjects factor.
Procedure. In a study ostensibly about consumer reactions to a product, participants were asked to carefully review either an angular chocolate bar or a rounded chocolate bar (Web Appendix C), and then rate how sweet and bitter they thought the chocolate bar would taste (1 = not at all, 9 = very much). Participants then indicated their attitudes toward the product shapes with the same 3-item scale used in the previous experiments (α = .92). Next, as with Study 1A, participants were shown a series of slider line scales, anchored by angular and rounded shapes at the left (-4) and right end (4), which have been widely used in previous research (e.g., Velasco et
al. 2016; see Web Appendix C). The scale items consisted of four visual scale items anchored with different angular and rounded shapes, and one semantic scale item anchored with the words angular and rounded. Participants were asked to drag the slider along the scale to a point that best matched bitterness or sweetness. Order of the items was counterbalanced. Composite measures of the mappings between shapes and tastes were computed by averaging the scores on the five scale items, yielding composite measures for sweet and for bitter tastes.
Results and Discussion
An independent samples t-test confirmed that participants’ liking for the product shapes did not differ between the angular and rounded conditions (Mangular = 5.45, SD = 1.61, Mrounded = 5.64, SD = 1.87, t(98) = .55, p = .58).
Shape–taste crossmodal correspondences. Separate one-sample t-tests with 0 (the mid- point of the scale) as the test value showed that the association between angular shapes and bitterness (M = -2.36, SD = 1.35, t(99) = -17.42, p < .001), and between rounded shapes and sweetness (M = 2.36, SD = 1.34, t(99) = 17.54, p < .001) were both significant. The absolute values of these associations did not differ (t(99) = .001, p = .99), indicating similar associative strengths.
Shape–taste crossmodal effects. Given that the strength of the shape–taste associations did not differ for angular and rounded shapes (in fact, they were virtually identical), we expected that both the angular and rounded shapes would show similar effects on taste judgments (unlike Study 1A). To test these hypotheses, we conducted a 2 × 2 mixed-model ANOVA, with product shape as a between-subjects factor and taste judgment as a within-subjects factor. The main effect of taste judgment (F(1, 98) = 1055.32, p < .001) and the shape × taste judgment were
significant (F(1, 98) = 8.11, p = .005). Participants rated the chocolate bar as more bitter in the angular condition (Mangular = 1.94, SD = .93) than in the rounded condition (Mrounded = 1.54, SD
= .76, F(1, 98) = 5.50, p = .021), and rated the chocolate bar as sweeter in the rounded condition (Mrounded = 7.62, SD = 1.12) than in the angular condition (Mangular = 7.04, SD = 1.24, F(1, 98) = 5.99, p = .016). Thus, angular shapes enhanced expectation of the chocolate’s bitterness and rounded shapes enhanced expectations of the chocolate’s sweetness.
Moderating effect of associative strength. We tested moderation with Hayes’ (2013) PROCESS Model 1. As expected, the interaction between product shape and associative strength between angular shape and bitterness (B = -.28, SE = .12, t = -2.29, p = .02) and between
rounded shape and sweetness (B = .37, SE = .18, t = -2.07, p = .04) were significant, indicating that the associative strengths moderate the crossmodal effects of shapes on tastes. A floodlight analysis indicates that for the angular shape–bitterness correspondence, the Johnson-Neyman threshold point (application threshold) at which the effects become significant is an associative strength of -2.16, and for the rounded shape–sweetness correspondence, the threshold is 1.97.
Thus, at an aggregate level, given that participants’ associative strength of angular shape and bitterness (M = -2.36) was above the activation threshold (-2.16), angular shapes enhanced expected bitterness of the chocolate bar. Similarly, the associative strength of rounded shape and sweetness (M = 2.36) was above the activation threshold (1.97), and thus rounded shapes
enhanced expected sweetness of the chocolate bar.
The results of Study 1C replicate the findings of the previous two studies, but with a different product category and associated tastes. Angular shapes were associated with bitter chocolate tastes, whereas rounded shapes were associated with sweet chocolate tastes. Unlike the previous studies, however, the strength of these associations did not differ. Consequently, their
downstream effects on taste judgments also did not differ. Moreover, in both cases, associative strength moderated these downstream effects, but because the strength of association was above the activation threshold for both angular and rounded associations, both angular and rounded shapes significantly affected taste judgments of bitterness and sweetness, respectively.
STUDY 2: AGE DIFFERENCES IN SHAPE–TASTE EFFECTS (CHILDREN) Study 2 tested age as a naturally occurring boundary condition for both shape–taste crossmodal associations and their downstream effects. If the crossmodal correspondences are learned, then the existence and associative strengths of the correspondences should increase with age. More precisely, we expect that shape-taste crossmodal associations are unlikely to be observed for children in the perceptual stage of cognitive development (roughly, < 7 yrs.),
because they have not yet developed the ability to associate abstract concepts, but are likely to be observed in children who have entered the analytical stage (> 7 yrs.). If so, then not only should this age threshold determine the presence of shape–taste correspondences, it should also
determine whether the correspondences influence downstream judgments, as shown in the previous experiments with adults.
Method
Participants and design. Eighty-one children recruited from a primary school in China (43 girls, range: 6-14 years, Mage = 9.81 years) participated in the study. Sample size was dependent on how many children were in school and able to participate; all who wanted to participate did. The participants were randomly assigned to conditions in a 2 (product shape:
angular, rounded) × 2 (taste judgment: bitter, sweet) mixed design, with product shape as a between-subjects factor and taste judgment as a within-subjects factor.
Procedure. Participants were asked to carefully view either an angular or a rounded chocolate bar (see Web Appendix D), then rate how sweet and bitter they thought the chocolate bar would taste (1 = not at all, 9 = very much), and indicate their attitudes toward the product shapes along the same 3-item scale (α = .78) used in the previous studies. Participants then completed the same slider scale measures used in Study 1C, but with only the four visual shape stimuli (Web Appendix D).
Results and Discussion
An independent samples t-test confirmed that participants’ liking for the product shapes did not differ between the angular and rounded conditions (Mangular = 5.84, SD = 2.27, Mrounded = 5.84, SD = 2.16, t(79) = .009, p = .99).
Age-related differences in shape–taste correspondences. We first examined shape–taste correspondences as a function of age. We expected that the correspondences would develop over time (with age), and thus age should be positively correlated with the magnitude of shape–taste correspondences. OLS regression showed that the correlations were significant for the angular–
bitter correspondence (B = -.23, SE = .10, t = -2.23, p = .029) and the rounded–sweet correspondence (B = .37, SE = .08, t = 4.74, p < .001).
To further examine the nature of this relation, we divided the age range (6-14 years) into three segments based on the standard deviations: younger children (< 7.85 yrs., < 1 SD), older children (> 11.77 yrs, > 1 SD), and middle children (7.85 - 11.77 yrs.), which correspond roughly to the age ranges for John’s (1999) stages of cognitive development (perceptual, analytical,
reflective). We then conducted separate mixed-model ANOVAs, with age group as a between- subjects factor and angular–bitter or rounded–sweet associative strengths as within-subjects factors.
For the angular shape–bitterness correspondence, only the main effect of shape–taste associative strength (F(1, 78) = 9.34, p = .003) was significant, but the main effect of age group (F(2, 78) = 2.09, p = .13) and the interaction (F(2, 78) = 2.09, p = .13) were not. Younger children did not show a significant angular–bitterness association (M = -.06, SD = 2.03, F(1, 78)
= .02 p = .90), whereas middle (M = -.70, SD = 1.82, F(1, 78) = 7.18, p < .001) and older children (M = -1.43, SD = 1.79, F(1, 78) = 8.86, p = .004) did.
For the rounded shape–sweetness correspondence, the main effects of shape–taste associative strength (F(1, 78) = 29.11, p < .001), age group (F(2, 78) = 3.92, p = .024), and their interaction (F(2, 78) = 3.92, p = .024) were significant. Younger children did not show a
significant rounded–sweet association (M = .24, SD = 1.67, F(1, 78) = .39, p = .54), whereas middle (M = 1.28, SD = 1.65, F(1, 78) = 33.27, p < .001) and older children (M = 1.75, SD = 1.12, F(1, 78) = 18.58, p < .001) did. Thus, similar patterns of effects as a function of age are observed for both angular–bitter and rounded–smooth correspondences, but the correspondence is somewhat stronger for the rounded–sweet correspondence.
Age-related differences in shape–taste crossmodal effects. We next tested whether the effects of the shape–taste correspondences on downstream taste judgments were moderated by age by conducting separate moderation analyses via Hayes’ (2013) PROCESS Model 1, with product shape (angular vs. rounded) as independent variables, taste judgment (bitterness or sweetness) as dependent variables, and age as a moderator. For the effect of angular shape on bitter tastes, the product shape × age interaction was marginally significant (B = .34, SE = .18, t =
1.87, p = .065). The angular shape did not affect bitterness judgments for younger children (B = -.07, SE = .50, t = -.15, p = .89), consistent with the finding that the angular–bitterness
association did not reach the activation threshold. In contrast, the effect for the middle age group was marginally significant (B = .59, SE = .35, t = 1.68, p = .10), and the effect for older children was significant (B = 1.25, SE = .50, t = 2.52, p = .01). The Johnson-Neyman technique showed that the application threshold at which angular shape affects bitterness perceptions is 10.17 years of age.
For the effect of rounded shape on sweetness taste judgments, the product shape × age interaction was significant (B = -.32, SE = .12, t = -2.68, p = .009). The rounded shape did not affect sweetness judgments for younger children (B = -.17, SE = .33, t = -.51, p = .61), whereas the effects for the middle age group (B = -.78, SE = .23, t = -3.42, p = .001) and older age group (B = -1.40, SE = .32, t = -4.32, p < .001) were significant. The Johnson-Neyman technique showed that the application threshold at which rounded shape affects sweetness perceptions is 8.92 years of age.
These results suggest that shape–taste crossmodal effects influence children’s taste judgments, but only for children who are at the analytical and reflective stages of cognitive development. The results also suggest that shape–taste application thresholds do not necessarily occur at the same age. The angular shape–bitterness application threshold occurs approximately 15 months later than the rounded shape–sweetness threshold. Thus, children appear to be able to abstract the common features of rounded shapes and sweetness earlier than they are able to abstract the common features of angular shapes and bitterness.
The results of Study 2 replicate the general patterns of results observed in the previous studies, but only for children who have reached a sufficient cognitive development stage. The
findings further bolster our theoretical reasoning, but also have implications for marketing applications for children, which we discuss in more detail presently. In the next two studies, we further investigate the processes that underlie the crossmodal effects on downstream judgments.
STUDY 3A: VISUALIZATION-INDEPENDENT SHAPE–TASTE CROSSMODAL EFFECTS Study 3A tested whether visualization is a necessary process in the generation of shape–
taste crossmodal effects. We manipulated participants’ abilities to generate a mental image via a visual load manipulation. Because the generation of mental imagery requires the involvement of visuospatial working memory (Baddeley and Andrade 2000), if a visualization process is
necessary, then inhibiting visuospatial working memory should reduce or eliminate the crossmodal effects of shape on taste judgments. We also included a traditional cognitive load manipulation that inhibits phonological working memory to rule out the possibility that mental load in general is driving the effects, rather than being specific to visualization. To simplify, we only included sparkling water as the product type.
Method
266 French college students (175 women; Mage = 21.99 years) participated in the study in return for the possibility of winning a lottery paying 30€. Participants were randomly assigned to conditions in a 2 (logo shape: angular, rounded) × 3 (mental load: visual, cognitive, no load) × 2 (taste judgment: carbonation, smoothness) mixed design, with logo shape and mental load as between-subjects factors and taste judgment as a within-subjects factor. Participants were informed that they would be completing two separate, unrelated tasks. The first task constituted the manipulation of either visual or cognitive load, both well-established in the literature.
